Room: Osceola Ballroom C
Purpose: Determine the feasibility of using TPS built-in registration assessment scripting tools, such as Dice and Image Similarity Coefficients (ISC) and distance to agreement (DTA). We assessed the quality of in-house procedures used for motion assessment of both gantry-mounted linac and Cyberknife units.
Methods: Python scripts generate ITVs for linac-based treatment plans and provide tracking and margin recommendations for Cyberknife SBRT. Motion assessments are performed using Hybrid intensity and structure based deformable registrations to either generate an ITV or confirm that fiducial motion tracks with target motion. Target contours are first drawn on the planning CT and mapped to the individual phases of the 4DCT series. Accurate deformable registrations are critical for the success of this method, since target motion is evaluated based on the results of these mapped structures. Magnitude of target motion and, in the case of Cyberknife, the consistency of motion between fiducials and the target, are then examined via scripted reporting of the Dice and ISC.
Results: The ISC can range from 0 to 1, with 0 representing no agreement and 1 representing perfect pixel agreement for a given Region-Of-Interest (ROI). For the clinical cases examined, we report a range of 0.5 to 0.96. An ISC value of 0.8 or greater indicates a good fusion, resulting in negligible structure mapping error. Dice similarity coefficient and DTA show the volume of ROI overlap and magnitude of motion when comparing rigidly copied to deformed structures. Max DTA is used for tracking recommendation on Cyberknife plans (Synchrony versus non-Synchrony).
Conclusion: ITV(s) and tracking recommendations for motion-management can be generated automatically based on motion assessment of the 4DCT. A script to automate this process is feasible, and can include additional margin recommendations to account for deformable fusion error as well as contouring errors generated by gaps between 4DCT phase data.